Teardown: A North-Facing Windowsill Tray at Four Years

The original build

A shallow glazed stoneware tray, thirty-two centimetres by eighteen, forty millimetres deep, with no drainage hole. A one-centimetre base of coarse pumice, a fibre separation layer, then two centimetres of a lean mix: composted bark, coarse sand, a little leaf mould.

Planted with Leucobryum cushions on the raised left third, a Hypnum carpet across the centre, a small drift of Bryum along the exposed right edge, and two young ferns — a maidenhair spleenwort and a soft shield fern seedling — set against a piece of weathered granite.

Position: a north-facing kitchen sill in a temperate maritime city. Measured light ranged from about 900 lux on a December afternoon to 4,500 lux on a bright June morning. No supplementary lighting was ever added.

Year one: the losses

The maidenhair spleenwort died in the first February. Post-mortem was straightforward: an open tray on a sill above a radiator, in a room dropping to fourteen degrees at night with dry heated air by day. It was the wrong plant for that microclimate rather than the wrong plant for the tray.

The Bryum drift lost its silvery compactness within four months, stretching and dulling under light that was adequate for cushion moss and far too weak for a pavement species. It persisted as an unremarkable green fringe for another year, then was outcompeted.

The Hypnum, by contrast, doubled its coverage in eight months and began climbing the granite. The Leucobryum held its shape but grew barely at all — normal for the species, and a useful reminder that a static cushion is not necessarily an unhealthy one.

Years two and three: equilibrium

By month eighteen the tray had settled into a stable pattern: Hypnum occupying roughly seventy per cent of the surface, Leucobryum holding the raised corner, the shield fern producing three or four new fronds a year and never exceeding twelve centimetres.

Watering settled at a thorough mist every three days in summer, weekly in winter, using rainwater collected from a garden butt. The tray was rotated 180 degrees monthly, which visibly reduced the lean of the fern toward the glass.

Two interventions in this period: a thinning of the Hypnum in month twenty-two, lifting about a third in irregular patches, and removal of a liverwort that arrived as a hitchhiker and began smothering the Leucobryum margin. Both were straightforward because the tray is open and everything is reachable.

What the teardown found

The pumice layer had partially filled with fine material washed down over four years, reducing but not eliminating its function. There was no sour smell anywhere in the tray and no black anaerobic zone, which for a vessel without a drainage hole was the most reassuring finding.

Substrate depth had dropped from twenty millimetres to about fourteen through decomposition and compaction. The upper five millimetres had become a dense mat of moss rhizoids and fine root, effectively a living crust that held together in one sheet when lifted.

Invertebrates were abundant and unintroduced: a large springtail population, several species of mite, two small earthworms presumably arriving in the leaf mould, and no sign of the fungus gnats that plagued months three to six. Nothing had ever been added deliberately.

Four things the maker would change

First, no radiator sill. The single greatest source of stress across four years was the daily temperature and humidity swing, and it cost one plant outright and slowed everything else.

Second, drop the Bryum. Mixing a full-sun pavement species with shade cushion mosses in one tray guaranteed one of them would be in the wrong conditions, and the light level was never going to satisfy both.

Third, more depth at the back. Forty millimetres of tray gave almost no scope for gradient, and the composition stayed flat throughout its life. Sixty millimetres at the rear would have transformed the layout for very little extra cost.

Fourth, plan the thinning from the start. The Hypnum's growth rate was entirely predictable, and building the layout around a species that would need managing three times a year — rather than reacting when it closed up — would have kept the design legible far longer.

Extended notes

• Total intervention over four years came to roughly nine hours, most of it watering. The two thinning sessions accounted for about ninety minutes.
• The granite was reused in the replacement build. Weathered stone that has already grown a moss film is worth far more than new stone.
• Cost of the original build was modest and dominated by the tray itself. The plants were divisions and a single nursery fern.

Frequently asked